HYDROCARBON FORMATION AND OXIDATION

IN SPARK-IGNITION ENGINES

Final Report on a Research Program

Funded by General Motors Research Laboratories

September 1976 to August 1979

Volume I: SUMMARY

John B. Heywood,* James C. Keck,t and Joe . Rifet

Sloan Automotive Laboratory

The Energy Laboratory

Massachusetts Institute of Technology

January 1980

(� 11 - -

*Professor, Department of Mechanical EngineeringtFord Professor of Engineering, Department of MechanicalEngineeringfPrincipal Research Engineer, Energy Laboratory, andLecturer, Department of Mechanical Engineering

ABSTRACT

This report summarizes the key results and conceptual findings from

a three year research program on hydrocarbon formation and oxidation

mechanisms in spark-ignition engines. Research was carried out in four

areas: laminar flame quenching experimental and analytical studies;

quench layer studies in a spark-ignition engine using a rapid-acting

gas sampling valve; flow visualization studies in a transparent engine

to determine quench layer and quench crevice gas motion; studies of heat

transfer, mixing and HC oxidation in the exhaust port. More detailed

descriptions of the individual research activities in these areas can

be found in the theses and publications completed to date which form

Volumes II to XI of the final report on this program.

-i-

TABLE OF CONTENTS

Page

BACKGROUND. . . .. . . .. . . . . . . . .

LAMINAR FLAME QUENCHING STUDIES. . . . . . . . . ..

ENGINE QUENCH LAYER STUDIES ...............

FLOW VISUALIZATION STUDIES ................

EXHAUST PORT STUDIES ...................

SUMMARY ........................

Laminary Quenching Studies. .. . . . . . . .. .

Quench Layer Studies in an Engine . . . . . .. .

Flow Visualization Studies in the Transparent Engine

Studies of Heat Transfer, Mixing and HC Oxidationin the Exhaust Port. . . . . . . . . . . . . . . . .

ACKNOWLEDGMENT . . . . . . . . .... .. .

APPENDIX. ........................

-ii-

1

6

11

16

23

29

29

29

30

30

32

33

. . .

. . .

. . .

. . 0

Volume II

"Standoff Distances on a Flat Flame Burner", by C. R. Ferguson andJ. C. Keck

Volume III

"Laminar Burning Velocity of Mixtures of Air with Indolene, Isooctane,Methanol and Propane", by M. Metchalchi and J. C. Keck

Volume IV

"Experimental Study of Quench Layer Hydrocarbons During the ExpansionStroke of a Spark Ignition Engine, Using a Fast Sampling Valve", byA. K. Wrobel, J. C. Keck and W. A. Woods

Volume V

"A Parametric Study of Quench Layer Properties by Rapid Sampling in aSpark Ignition Engine", by P. Weiss and J. C. Keck

Volume VI

"The Design and Testing of a Flow Visualization Engine", by S. P. Hansenand J. M. Rife

Volume VII

"Schlieren Flow Visualization of Combustion and Quench Zone Motion in anInternal Combustion Engine", by J. Sanchez Barsse, J. B. Heywood andJ. M. Rife

Volume VIII

"Schlieren Visualization of the Flow and Density Fields in the Cylinderof a Spark-Ignition Engine", by M. Namazian, S. Hansen, E. Lyford-Pike,J. Sanchez-Barsse, J. Heywood and J. Rife

Volume IX

"Heat Transfer, Mixing and Hydrocarbon Oxidation in an Engine ExhaustPort", by J. A. Caton and J. B. Heywood

Volume X

"Visualization of the Flow of Exhaust Through an Internal CombustionEngine Exhaust Port", by J. W. Harrington and J. M. Rife

-iii-

Volume XI

"Quenching Studies to Determine Hydrocarbon Oxidation in the ExhaustPort of a Spark-Ignition Engine", by J. V. Mendillo and J. B. Heywood

-iv-

1. BACKGROUND

From September 1976 through August 1979, the Engine Research

Department at the General Motors Research Laboratories supported an

extensive fundamental research program on hydrocarbon formation and oxi-

dation in spark-ignition engines in the Sloan Automotive Laboratory at

M.I.T. Three Ph.D. theses, six Masterts degree theses and one Bachelor's

degree thesis have been completed under this program; one Ph.D thesis and

one Master's degree thesis initiated under the program are still in pro-

gress. This summary report describes the major findings of this research,

and explains the importance of each of these pieces of research to devel-

oping our understanding of the overall hydrocarbon formation mechanism.

The theses and publications completed to date on this program are listed

in Appendix A. These were submitted to General Motors during the course

of the program as they became available. To provide a complete record of

this program, they are submitted with this summary as Volumes II through

XI of the final report.

The study of spark-ignition engine hydrocarbon emissions mechanisms

continues to be an important research problem. The latest hydrocarbon

emission standards introduced in 1980 have brought with them a fuel economy

penalty, which is likely to be made worse by the stricter 1981 NO standard.

Evermore stringent fuel economy standards demand the most aggressive engine

design and calibration strategy possible. Thus hydrocarbon emission con-

trol is likely to continue as an ever present constraint. Though the use

of catalyst technology is now the major HC control strategy, materials

1

availability and pressures to reduce emission system cost, and in the

longer term changing fuel composition, will insure that control of engine

HC emissions will remain an important research and development activity.

Our conceptual understanding of the total hydrocarbon emissions for-

mation mechanism is far from adequate. Prior to this current research

program, the major insights into the basic mechanisms had come from work

by Daniel and Wentworth at General Motors, who demonstrated the existence

of quench layers in a spark-ignition engine, the importance of the crevice

between the piston and cylinder wall, the influence of wall temperature

and surface roughness, and the non-uniformity in hydrocarbon concentration

within the engine exhaust gases. Work in the Sloan Automotive Laboratory

at M.I.T. had shown that about half the mass of hydrocarbon emissions left

the cylinder during the exhaust blowdown phase, and about half left the

cylinder at the end of the exhaust stroke. It was hypothesized that the

blowdown hydrocarbons came from quench layers on the cylinder head and

upper part of the cylinder liner. It had been shown in water analog flow

visualization studies that during the exhaust stroke the piston scraped

the boundary layer fluid off the cylinder wall and rolled it into a vortex;

it was suggested that this vortex could be the mechanism by which hydro-

carbons quenched in the crevice between the piston and the cylinder wall

eventually left the cylinder. Thus, at the start of this program, the

major elements in a qualitative conceptual model of the total hydrocarbon

emission mechanism had been proposed.

The sequence of steps which make up this complete hydrocarbon emis-

sions formation and oxidation model is extremely complicated. The flame

2

quenches at the walls of the combustion chamber, leaving a quench layer

of unburned and partially burned fuel-air mixture adjacent to the wall.

The flame also quenches at the entrance to any crevices where the opening

is too small to permit the flame to enter. During the expansion process,

mixing at the outer edge of the quench layers is expected to occur as well

as some oxidation of the unburned fuel species. As the cylinder pressure

falls, gas will flow out of the crevices into the cylinder where mixing

with the bulk cylinder gases and some oxidation of the unburned mixture

would be expected. During the blowdown phase of the exhaust process, the

rapidly exiting gas would entrain quenched gases from the boundary layers

on the cylinder head and upper cylinder wall. During the exhaust stroke

the piston scapes the cylinder walls of any gases laid there during expan-

sion and pushes a large fraction of the hydrocarbons remaining in the

cylinder after blowdown into the exhaust. Within the exhaust port, exhaust

manifold and exhaust pipe, depending on the local oxygen concentration and

temperature, additional oxidation of unburned hydrocarbons is expected to

occur. There is considerable evidence to support this qualitative description.

The primary goal of the research program summarized here was to test,

refine and quantify where possible these conceptual ideas and provide a

more coherent model of the overall process. The following are key areas

where data and predictive models have been lacking:

1) the effects of unburned mixture conditions and compositionon the flame quenching process at the combustion chamber walls;

2) the effects of wall geometry, wall temperature, and thermalboundary layer structure on this flame quenching process;

3) the effect of fuel composition on the quench layer thicknessand its contents;

3

4) the details of the quench crevice hydrocarbon generationmechanism and the influence of piston and piston ring geometry;

5) the motion of quench layer and crevice gas during the expansionand exhaust processes;

6) the oxidation of quench zone hydrocarbons subsequent to theirformation, within the cylinder during the expansion and exhaustprocesses;

7) data which separates the in-cylinder formation and oxidationprocesses from mixing and oxidation processes in the exhaustport, manifold and tail pipe; and

8) the importance of hydrocarbon oxidation in the exhaust system asa function of engine operating conditions and exhaust systemdesign.

The research carried out at M.I.T. and summarized in this report has

contribued towards resolving several of these areas of uncertainty in two

ways--first, by providing experimental data at selected intermediate stages

in the overall hydrocarbon formation process which help define the role of

the individual subprocesses in the total mechanism, and second, by developing

quantitative scaling laws for key subprocesses where sufficient experimental

data already existed or could be obtained. Work has been focused in four

distinct areas:

1) Laminar flame quenching studies;

2) Engine quench layer studies with a rapid acting sampling valve;

3) Flow visualization in a unique square cross section transparentengine;

4) Studies of heat transfer, mixing and hydrocarbon oxidation inthe engine exhaust port.

The laminar quenching studies were carried out to define the flame

quenching process in a simple geometry system, where precise experimental

data and an analytic theory could be developed. The engine quench layer

4

studies had the goal of measuring quench layer properties in an operating

engine, providing data on the variation of quench layer properties during

the engine operating cycle as a function of major engine operation variables,

and providing a test for the laminar flame quenching theory in the engine

environment. The flow visualization studies were initiated to provide

information on the motion of gas originating in quench regions, during the

expansion and exhaust processes. Both qualitative and quantitative infor-

mation on these flow processes was sought; this was regarded as a key area

where conceptual ideas needed to be developed and substantiated. The

exhaust port studies had the dual goal of separating in-cylinder and exhaust

system mechanisms, and developing and validating a model which could be used

to optimize exhaust system design. The major results in each of these four

research areas will now be reviewed.

5

2. LAMINAR FLAME QUENCHING STUDIES

Although the combustion process in a spark-ignition engine is turbu-

lent, we concluded that a laminar flame quenching study was relevant to the

engine context and had substantial advantages as a starting point for a

flame quenching analysis. In an engine, as the flame approaches the wall,

it propagates into a viscous sublayer adjacent to the wall where turbulent

fluctuations have negligible energy. Because the wall remains cold, the

flame is extinguished a small distance away from the wall leaving unreacted

fuel and carbon monoxide. The quenching distance is so small (50-500 m)

that it is extremely difficult to study the process directly in the engine

environment. A simpler geometry steady-flow system was chosen for study:

the standoff of a laminar flame above a heat sink in a one-dimensional

laminar premixed-flow. It was hypothesized that the standoff distance in

this problem was analogous to the quench layer thickness in an engine.

The details of this work are given in Volume II.

The problem for which a theoretical analysis was developed was a

one-dimensional laminar flame with heat loss to an upstream cold boundary.

The analysis relates the flame speed, the burned gas temperature and the

standoff distance of the flame above the cold boundary. It was shown that

there is both a high temperature and a low temperature solution for the

standoff distance. Also, there is a minimum distance for which solutions

exist. This minimum standoff distance is the quench distance for a tran-

sient flame propagating towards a wall, when the quasi-steady approximation

6

is valid. It was shown that a Peclet number based on laminar flame speed

and standoff distance depends only on the ratio of the heat of combustion

to the heat loss to the cold wall provided that: (i) the dimensionless

activation energy B = E/RTb is greater than about 7, and (ii) that when

hydrogen is a reactant it carries little of the sensible enthalpy. It was

then shown that because so little heat is required to quench a flame, the

minimum standoff distance is a good approximation to the quench distance

for an unsteady flow. An expression was derived for the mass of fuel,

per unit area of the wall, in a quench layer for use in engine quench

layer calculations.

Experiments were carried out in a laminar porous-metal flat-frame

burner to measure the flame standoff distance from the burner, as a func-

tion of fuel and unburnt mixture composition. A hot-wire pyrometer was

developed to measure flame temperature and thus locate the flame front

position relative to the porous-metal burner-surface. Acceptable quanti-

tative agreement between the standoff Peclet number based on the measured

standoff distance, as a function of the ratio of the heat of combustion to

the heat loss, and the limiting Peclet number derived from the theory was

obtained.

An analysis was also made of laminar flame quenching in flows parallel

to a heat sink. The fundamental question examined was the appropriate

characteristic length over which a flame in a tube loses heat. Two

The Peclet number is the ratio of convective to conduction heat transport.The standoff Peclet number used in this study was defined as Pe =(puSLCpD)/X where pu is the unburned gas density, SL is the laminar flame

speed, c the specific heat at constant pressure, D the standoff distance,

X the thermal conductivity.

7

possibilities were studied: the tube radius and the flame thickness. By

using a functional form prescribed by the analysis, a correlation of laminar

flame speed, quenching diameter, and lean limit flame temperature was

developed. Nearly two orders of magnitude of quenching diameter and flame

speeds were correlated by a Peclet number. The Peclet number is based on

the quench distance, flame speed, unburnt gas density, frozen specific

heat of the burned gas, and frozen thermal conductivity of the unburned

gas at the adiabatic flame temperature.

These theories were used to compute quench distances in an engine.

It was shown there is not much difference between the predictions with the

correlation for perpendicular flow and four-tenths of the predictions for

the parallel flow (0.4 is the accepted factor for relating the perpendicular

and parallel flow quenching distances). This is also not much difference

between using the free-stream unburnt-mixture temperature, or the wall

temperature, as the temperature characterizing the unburnt gas. Quench

layer thickness predictions were in approximate agreement with the engine

layer thicknesses in the literature determined by Daniel. It was con-

cluded that these laminar flame quenching models would therefore be

useful in the engine context.

A critical parameter in these laminar quenching theories is the

laminar burning velocity of the unburned mixture. Little reliable data

was available for the laminar burning velocity of mixtures of air with

fuels relevant to spark-ignition engines, under typical engine unburnt

mixture conditions. A facility for measuring the laminar burning velocity

of fuel-air residual gas mixtures had been constructed in the Energy and

Chemical Dynamics Laboratory which is associated with the Sloan Automotive

8

Laboratory at M.I.T. Accordingly, an extensive program of experiments to

develop laminar burning velocity data for mixtures of air with propane,

isooctane, methanol and indolene was initiated. This work is described in

detail in Volume III.

The facility consists of a spherical combustion bomb in an oven which

can be heated up to 500 K. The bomb is equipped with ionization probes to

measure the arrival time of the flame at the wall and check the spherical

symmetry of the flame front. A piezoelectric pressure transducer to measure

pressure during the combustion process, and a balanced pressure indicator to

calibrate the transducer are built into the bomb. A laser shadowgraph

system was used to measure the arrival time of the flame front at a known

radius, and also to check the assumption of negligible pre-flame reaction.

A thermodynamic model was developed and used to calculate the laminar

burning velocity from the pressure time history obtained during the combus-

tion process. The unburnt mixture properties were computed using therm-

dynamic data from the JANAF Tables, and the assumption of frozen compo-

sition; the burned gas properties were computed using an approximate

equilibrium model for the burned gases. Allowance was made for heat loss

to the bomb walls.

Laminar burning velocities for mixtures of propane-air, isooctane-air,

methanol-air, and a multi-component hydrocarbon fuel (indolene)-air were

measured in the pressure range 0.4 to 40 atm and temperature range 298 to

750 K, for fuel-air equivalence ratios from 0.8 to 1.5. Both power law

and exponential relations were fitted to this data to permit easy estima-

tion of laminar flame speed as a function of pressure, temperature and

9

fuel-air equivalence ratio. In addition, the effects of residual gases on

the laminar burning velocity of stoichiometric isooctane-air mixtures were

measured; the residual gases were simulated by mixtures of carbon dioxide

and nitrogen to obtain the appropriate specific heat.

Methanol-air mixtures were found to have the highest laminar burning

velocity amongst the fuel-air mixtures studied, followed by propane-air,

isooctane-air and indolene-air mixtures. The difference between the iso-

octane and indolene values was small, however. The laminar burning velocity

peaks at an equivalence ratio of about 1.1 and falls off rapidly as the

mixture becomes both richer and leaner.

The burning velocity correlations developed under this program are

based on a sufficiently extensive data set for us to conclude that this

variable is no longer a major area of uncertainty in quench layer corre-

lations. This data set and these correlations obviously have extensive

application in other areas of combustion. In particular, they are proving

extremely useful in spark-ignition engine flame propagation modeling.

10

3. ENGINE QUENCH LAYER STUDIES

To define the contribution from quench layers on the cylinder head

and upper portion of the cylinder liner, to exhaust hydrocarbons, a program

of engine experiments was undertaken. The goal of this part of the research

program was to sample gases (including gases from the quench layer) after

combustion, from the cylinder using a rapid-acting sample valve, and through

use of a model for the flow into the valve develop more detailed information

about the quench layer characteristics than has been available to date. By

carrying out these experiments over a range of engine operating conditions,

the variation of quench layer thickness and hydrocarbon content could be

determined. Volumes IV and V describe the results of this activity.

For the engine experiments, a rapid-acting sampling valve supplied

to the Sloan Automotive Laboratory by General Motors Research Laboratories

was modified and inserted through the top of the cylinder liner in a single

cylinder CFR engine. This sample valve, which had an open time of about

two milliseconds, was used to extract a gas sample once per engine operating

cycle at a pre-selected crank angle. A model was developed for a flow into

the valve during the sample period, and was used to interpret the results

of gas concentration measurements made on the sample valve flow. This

model permitted an estimate to be made of mass of hydrocarbons in the

quench layer, per unit wall area, and of the quench layer thickness.

The basis for the model of the flow into the sample valve was the

following. The mass of hydrocarbons sampled per cycle is computed from

the hydrocarbon concentration in the sample stream and the sample flow rate.

11

The mass of hydrocarbons sampled per cycle was then corrected for leakage

which occurs while the valve is nominally closed. The leak rate with the

valve closed was measured. The mass of hydrocarbons leaked was computed

by assuming the leak rate to be proportional to the pressure difference

between the cylinder and the valve sample line, and that prior to time-

of-quench, the leaked gas is unburnt mixture while after time-of-quench

the leaked gas is burned mixture with negligible hydrocarbon concentration.

It was found necessary to modify the design of the valve and use a copper

seat to keep the leakage correction to less than twenty percent. As a

first approximation the flow into the sample valve was assumed radial and

inviscid. The sample gas was assumed to occupy a hemispherical region

concentric with the sample valve orifice. The mass of hydrocarbons per

unit area on the cylinder wall (neglecting boundary layer effects) could

then be computed. A viscous model for the flow into the sample valve was

then developed, because it was clear that boundary layer behavior during

portions of the engine cycle would significantly retard the entrainment of

hydrocarbons on the wall. The axisymmetric boundary layer equations for

flow into a point sink on a plane surface were solved, followed by the

integration of an assumed quench layer profile over a volume specified by

the flow field. It was found that the ratio of the thickness of the quench

layer relative to the thickness of the viscous boundary layer is much

larger during exhaust than during expansion, so that exhaust stroke esti-

mates of quench layer hydrocarbons are likely to be much more accurate.

Functions which relate the mass of hydrocarbons per unit area calculated

from the inviscid model to the mass per unit area with the viscous model

12

were developed and computed for an appropriate range of engine conditions.

Two studies and sets of experiments were carried out. The first (by

Wrobel, Keck and Woods; Vol. IV) developed the gas sampling methodology

and model for flow into the valve, and investigated the effects of volume

sampled per cycle by varying needle lift and sample duration, and time of

sampling during the cycle. The second (by Weiss and Keck; Vol. V) improved

the experimental technique, explored through a set of engine experiments

the effect of variations in engine operating conditions on quench layer

properties, set up the laminar flame quenching model of Ferguson and Keck

(Vol. II) for predictions of quench layer properties in an engine, and

compared predictions made with this quenching theory with the experimental

engine data.

It was shown that the mass of hydrocarbons quenched per unit wall area

was independent of the volume sampled, at constant sample duration, and only

varied slightly with needle lift. In the studies where sampling occurred

at different times during the engine cycle, it was found that the mass of

hydrocarbons per unit area in the quench layer generally decreased during

the expansion stroke. Peak value determined using the sample valve flow

model described above occurred at about 180 degrees crank angle, during the

exhaust blowdown process. The high gas velocities within the cylinder

during blowdown violated the assumption made in the model that the

radial flow velocity toward the valve orifice was the dominant velocity

in the flow field; thus the model is not appropriate during blowdown.

During the exhaust stroke, the mass of hydrocarbons per unit area remain

essentially constant at the sample valve location used in this study

13

(close to the top of the cylinder wall).

Quench layer properties were determined in the CFR engine as a func-

tion of equivalence ratio, inlet manifold pressure, spark advance, percent

exhaust gas recycle and coolant water temperature. The values of mass of

hydrocarbons per unit wall area in the quench layer were relatively insen-

sitive to operating conditions and varied between about 0.4 and 1 g HC/cm2

The values of the quench layer thickness estimated from the mass per unit

area data ranged from. 0.05 mm early in the expansion process to 0.35 mm

during the exhaust stroke. The mass of hydrocarbons per unit area showed

a minimum at an equivalence ratio of 1.0, and showed a slight decrease with

increasing inlet pressure, increasing spark retard and cylinder wall tem-

perature, and a slight increase with increasing percentage EGR.

The laminar flame quenching theory described in the previous section

was used to provide a theoretical estimate of quench layer properties to

compare with the above experimental results. The laminar flame speed data

of Metghalchi (Vol. III) were used in the Peclet number correlation which

defines the minimum standoff distance, which was equated with the quench

distance as explained previously. The predicted mass of hydrocarbons per

unit area were comparable to the measurements, though higher by a factor

of between 1.1 and about 2, and generally showed the same trends with the

engine variables studied. Since entrainment from the quench layer during

blowdown (see next section) is likely to have reduced the mass of the per

unit wall area in the experimental results (which were evaluated during the

exhaust stroke) the theoretical predictions would be expected to lie above

the experimental data.

14

These results for mass of hydrocarbons in the quench layer, per unit

wall area, have been used to estimate the contribution which quench layers

on the cylinder head and upper portion of the cylinder liner could make to

total exhaust hydrocarbons. Based on a mass of hydrocarbons per unit

wall area of 1 ig HC/cm2, and a quench layer area swept into the exhaust

during blowdown equal in magnitude to the piston area, the quench layer

contribution would be of order 300 ppm C at wide-open-throttle and 600

ppm C at light load. These exhaust levels are small relative to CFR engine

exhaust hydrocarbon levels, but are a significant fraction of exhaust

levels typical of modern production engines.

The engine experiments in this program were carried out with rela-

tively clean combustion chamber walls. The fuel used was isooctane, and

deposit build up with this fuel under typical research engine usage is

not significant. The effect of deposits on these surfaces, on the flame

quenching process and quench layer properties has yet to be determined.

Also, for extremely lean or dilute engine operating conditions, the quench

layer contributions will be more substantial.

This study showed that the laminar quench correlations predict

quench layer properties with reasonable accuracy. Thus, the flame quench-

ing process with clean combustion chamber walls has, we believe, now been

defined with sufficient accuracy, both experimentally and theoretically,

for incorporation into an engine hydrocarbon model.

15

4. FLOW VISUALIZATION STUDIES

A substantial part of the effort in this program was devoted to the

design, construction and use of a transparent engine to visualize the flow

field within the engine cylinder from the time of flame quenching at the

combustion chamber walls to the end of the exhaust stroke. Through quali-

tative descriptions of how unburnt hydrocarbons moved from quench layers

and quench crevices at the outer edges of the combustion chamber to the

exhaust valve had been proposed, there has been no direct confirmation

of these conceptual ideas to date. It was anticipated that the motion

of the thermal boundary layers on the combustion chamber walls, which

contain the quench layers, and the flow out of crevices, during the

expansion and exhaust strokes, could be observed by techniques such as

Schlieren photography which identify regions where density gradients exist.

A square cross-section transparent engine was constructed for this

purpose. The engine was built on a CFR engine crankcase using the CFR

piston and cylinder as a crosshead for the square cross-section piston and

cylinder assembly. A square cross-section cylinder assembly was chosen to

permit optical access to the entire cylinder volume over the complete

engine operating cycle. The square cross-section assembly had two parallel

steel walls and two parallel quartz glass walls. The square cross-section

piston was fitted with three sets of hard graphite rings, dovetailed at

the corners and spring loaded, to seal against gas leakage. The CFR head

and valve mechanism completed the assembly.

16

It was shown through a thermodynamic analysis of engine pressure

data that the engine operates satisfactorily with propane fuel under

typical engine operating conditions. The combustion efficiency deter-

mined from this analysis, and the cyclic variation in peak pressure are

comparable in magnitude to conventional engine values. Spark-discharge

short-time exposure photographs, and high-speed movies, were taken while

the engine was operating, to define details of the flow and density fields

in the engine cylinder throughout the engine cycle. A complete description

of the engine, its operating characteristics, and an extensive set of

photographs taken in this facility, with an interpretation of those fea-

tures of the flow which were observable in the photographs and movies,

are given in Volumes VI, VII and VIII. Those flow features which relate

directly to the hydrocarbon formation mechanism are the following.

The turbulent flame front was observed to propagate into the thermal

boundary layer on the combustion chamber walls. The quench distance was,

as expected, too small to be observed directly. The thickness of the

dark denser-gas layer adjacent to the wall (interpreted as the thermal

boundary layer) did not change significantly from the time the front of

the flame arrived at the wall to the time the back of the flame reached

the wall. After time of peak cylinder pressure, early in the expansion

process, denser gas was observed flowing out of the crevice around the

spark plug electrode. Much later in the expansion process, gas was

observed to flow as a jet out of the piston top-land region and the volume

behind and between the piston rings. This piston crevice gas moved

relatively rapidly into the bulk of the cylinder. During the exhaust

blowdown process, the gas from the spark plug crevice exits the cylinder,

17

and a substantial portion of the boundary layer on the cylinder head and

upper cylinder liner walls is entrained in the blowdown flow. The gas

from the piston crevice reaches the exhaust valve during blowdown and

some of it exits the cylinder. During the exhaust stroke, the piston

pushes out much of this crevice gas with the bulk cylinder gases. As the

piston travels up the cylinder during the exhaust stroke, the piston

scrapes the boundary layer off the cylinder wall and rolls it into a

vortex in the piston-crown cylinder-wall corner. In the square cross-

section engine geometry, this vortex was not always stable. It was

observed to lift off the piston crown at the end of the exhaust stroke

and move towards the exhaust valve. Measurements of the size of this

vortex show that it grows during the exhaust stroke at a rate described

by the scaling law developed previously at M.I.T. by Tabaczynski and

others from water analog piston-cylinder studies.

A preliminary model was developed for gas flow into and out of the

crevice region between the piston and cylinder wall and behind the piston

rings. These volumes can be regarded as linked to the cylinder volume

through a set of flow restrictions. When the cylinder pressure is higher

than the pressure behind and between the rings, gas flows into this crevice

volume. During the expansion process, the pressure in the cylinder

eventually falls below the pressure in the volume between and behind

the rings, and the direction of gas flow reverses. Using volumes and

flow areas estimated from the geometry of the piston and ring assembly,

it was shown that this flow reversal should occur at about the point

when jets issuing from the crevice are observed in the Schlieren movies.

18

We have estimated that the volume behind and between the piston rings (in

the square cross-section engine) is about four times the equivalent volume

in a conventional engine, and the flow area connecting this volume with

the cylinder is also about four times that area in a conventional engine.

Since this preliminary model predicts behavior approximately consistant

with what was observed fron the Schlieren movies, a more detailed model

of the piston crevice volume, behind and between the rings, is under

development.

If the flow into and out of the region between the piston and cylin-

der wall in a conventional engine is similar to that observed in this

square cross-section engine, then our earlier suggestion that the vortex

in the piston-crown cylinder-wall corner would contain most of the piston

crevice gases would need to be modified. The jets of gas which issue from

this crevice region, once the flow reverses direction and is from the

crevice into the cylinder, may have sufficient velocity to penetrate a

substantial distance towards the exhaust valve before the blowdown

process is over. Note, however, that the gases in the crevice are likely

to come out in approximately the order in which they entered. The first

gas to enter the crevice, during the compression stroke and early part of

the combustion process before the flame reaches the crevice entrance,

should be unburned mixture. Following flame arrival at the crevice en-

trance, the gas entering the crevice will be burned gases until that

point in the expansion stroke when the flow reverses.

19

In summary, we have observed and described a new process by which

hydrocarbon rich gases from the crevice between piston, piston rings and

cylinder wall could leave the cylinder. The gases in this crevice enter

the cylinder as jets with substantial velocity. These jet flows provide

a possible, though as yet unsubstantiated, mechanism for crevice gases to

leave the cylinder during the blowdown process. At the same time, they

add to the vortex mechanism by providing hydrocarbon rich regions,

along the cylinder wall, above the piston, which would leave the cylinder

much later in the exhaust process.

From the Schlieren spark-photographs and high speed movies, other

features of the flow within the cylinder could be observed. These in-

cluded the jet type flow through the intake valve and the in-cylinder

motion this jet sets up, the turbulence scale and bulk motion during

compression, the ignition, flame growth and flame propogation processes,

details of the flame structure, examples of cycle-by-cycle variations,

and additional details of the bulk gas motion during exhaust. These

are described in more detail in Volume VIII.

To sum up this part of the total program, the use of a square

cross-section transparent side-wall spark-ignition engine for flow visual-

ization through the complete engine cycle has been successfully demon-

strated, and the engine operating characteristics were shown to be

sufficiently close to those of a conventional spark-ignition engine for

the results to provide useful insights into the flow fields and the

flame structure in real engines. Our observations and insights into

both the hydrocarbon emissions mechanisms, and other aspects of the flow

and combustion process can be summarized as follows:

20

1) For throttled operation, when the intake valve opens, thebackflow from the cylinder into the intake system is evidentand the bulk gas in the cylinder is drawn towards the valve.

2) The intake flow enters the cylinder as a connical jet whichtravels to and interacts with the cylinder wall.

3) A vortex type flow in the upper corners of the cylinder, setup by the intake jet, is visible in the Schlieren movies. Thisrotating motion in the upper corners persists through com-pression, combustion and expansion.

4) The turbulent flame which develops from the spark dischargepropagates as an approximately spherical though irregularfront. The flame zone is thick, the front to back distancebeing 10-15 mm, and a detailed internal flame structure isevident.

5) Only a fraction (of order one half during the earlier part ofthe burning process) of the mixture contained behind the pro-pagating flame front is fully burned.

6) The characteristic burning time of mixture within the flame isof order a few milliseconds (average 1.8 ms). This gives acharacteristic turbulent flame length scale of 14 mm (which iscomparable to the observed flame thickness), and a characteristiclaminar burning length scale of 1.3 mm.

7) The flame front shows both larger scale irregularities and asmaller scale structure. The larger scale irregularities areof order 10 mm in size (about 1/3 the clearance height). Thesmaller scale structure at the flame front appears to be burnedgas elements of order 2.5 mm in dimension surrounded by a thinflame zone and then unburned mixture. This smaller scale iscomparable to the laminar burning length scale. At the backside of the flame, the small scale structure is reversed:unburned mixture within a thin flame zone with burned mixtureoutside.

8) The spark discharge is observed as a thin column between theelectrodes which then grows as a flame kernel develops aroundthe spark. Once the kernel fills the electrode gap it developsirregularities. Once it extends beyond the electrodes (scale5 mm or greater) its appearance is similar to that of a fullydeveloped turbulent flame.

9) A comparison of pictures from different cycles shows differentdegrees of flame development at early times, and differentflame center motion and front shape. These initial differ-ences lead to substantial later differences in fully developed

21

flame front position and shape. The effect of cycle-by-cycle variations in mixture motion on the flame can beobserved. These differences in flame development and pro-pagation can be related to the value and time of occurenceof peak pressure.

10) During expansion (following time of peak cylinder pressure)flow of denser gas out of the spark plug crevice and theregion between the piston and the cylinder walls was observed.Gas from the crevice volume behind and between the rings issuedfrom the corners of the gap between the piston crown and thewalls as a jet, late in the expansion process, when thecylinder pressure fell below the pressure in this creviceregion. These crevices gases are expected to have a highunburned hydrocarbon concentration.

11) The blowdown flow was observed to entrain the denser thermalboundary layer gas off the cylinder head and upper part of thecylinder wall. Some of the gas from the crevice between thepiston crown and cylinder wall also leaves the cylinder duringblowdown.

12) During the exhaust stroke, the roll-up of a vortex in the cornerbetween the cylinder-wall piston-crown-surface was observed,though the vortex was not always stable in this engine geometry.The growth of this vortex fits the scaling law developed pre-viously by Tabaczynski, et al. in water analog piston-cylinderexperiments.

13) Dark layers, interpreted as thermal boundary layers, of thick-ness between 0.2 and 2.2 mm were observed on the walls of theengine throughout the cycle. The boundary layers on the cylinderhead and upper cylinder walls are thinnest when gas velocitiesare highest during intake and blowdown. The thickness grows toa maximum just before the exhaust valve opens.

The results obtained to date in the flow visualization engine at a

single engine operating condition and with one engine combustion chamber

geometry (square cross-section cylinder, uniform clearance height) have

thus provided substantial insights into both the flow aspects of the

hydrocarbon emissions mechanism, as well as details of the intake and

compression flows, and the combustion process. This facility has tremen-

dous potential for additional contributions in these areas.

22

5. EXHAUST PORT STUDIES

The motivation for devoting a significant fraction of the research

effort to studies related to hydrocarbon oxidation in the exhaust port

was the following. Engine experiments with exhaust port liners had indi-

cated that significant additional burn up of hydrocarbons could be

achieved through reductions in exhaust port heat transfer. Previous

exhaust system models had focused primarily on the manifold. Our pre-

liminary assessment indicated that most of the gas phase hydrocarbon

burn-up outside of the cylinder occurred in the exhaust port. Yet no

model was available, to our knowledge, for optimizing the design of the

port to maximize the reduction in hydrocarbons. A major thrust there-

fore was the development of an exhaust port flow model which included

heat transfer, mixing and hydrocarbon oxidation processes in adequate

detail, and the validation of that model through engine experiments. An

important additional reason for studying exhaust port processes was to

provide information which separates the mechanisms governing hydrocarbon

formation within the cylinder, from what happens to the hydrocarbons

which exit the cylinder within the exhaust system. A program of engine

experiments where the exhaust flow was quenched at selected locations

within the port was also planned and carried out.

The development of the exhaust port model, and its use, are des-

cribed in Volume IX. The modeling work first focused on heat transfer

processes within the port to enable the gas temperature as a function of

position and time to be determined. A series of heat transfer models

were developed for different phases of the exhaust process: e.g., a jet

23

impingment model for the valve stem and valve guide heat transfer for

small valve lift; a flow regime dominated by large scale fluid motion

set up by the jet-type flow exiting the exhaust valve, again for small

valve lift; a developing pipe flow model for the port during the high

valve lift phase of the exhaust process. Visualization studies of the

flow pattern in the exhaust port, described in Volume X, contributed to

the definition of these flow regimes.

To assist in developing these heat transfer correlations, experi-

ments were carried out in a CFR engine to measure the decrease in exhaust

gas temperature in the exhaust port, on a time and space resolved basis

during the exhaust process. The gas temperature at cylinder exit during

exhaust was estimated from cylinder pressure measurements using a ther-

modynamic model to determine the state of the gases within the cylinder.

To measure the gas temperature at exhaust port exit, a fine wire resis-

tance thermometer probe was developed, which could be traversed across

the exhaust-port exit-plane. Gas temperatures as a function of time were

determined with the resistance thermometer as a function of position at

port exit, engine load, speed, equivalence and spark advance.

The heat transfer models were coupled with a one-dimensional quasi-

steady, though time varying, model of the flow through the port, and

were developed to give good agreement with the experimentally determined

gas temperatures at port entrance and exit. A simple correlation based

on the gas velocity at the valve seat was found to give an adequate

estimate of the heat transfer in the port. A four time period heat

transfer correlation was also developed, based on two distinct flow

regimes for low and high valve lift, which provided the best overall

agreement with the measured gas-temperature data.

24

A mixing model was added to this flow model to approximate the

mixing which takes place between gas at the end of one exhaust pulse

and the beginning of the next pulse. An analysis of overall hydrocarbon.

oxidation rate expressions in the literature relevant to exhaust system

hydrocarbon oxidation identified a suitable oxidation rate expression

which was coupled with the flow, heat transfer and mixing models.

Several possible hydrocarbon concentration profiles at the exhaust port

inlet were tested in conjunction with the model to provide port exit

concentration profiles for comparison with available time-resolved

experimental hydrocarbon concentration data. A two level concentration

profile was found to give reasonable agreement. A sensitivity analysis

indicated the detailed shape of hydrocarbons concentration profile at port

inlet was not important to the fraction reacted in the port. An analysis

of various mixing volumes within the port showed that the precise choice

of mixing volume was not important either.

A model study of the effect of changes in key engine operating vari-

ables on the percentage reduction in cylinder exit hydrocarbons which

occurs in the exhaust port was then carried out. The base point for this

study was an indicated mean effective pressure of 345 kPa and speed of

1400 rev/min. The predicted percent hydrocarbon reacted in the port varied

between about ten and forty percent. Percent reacted decreased with in-

creasing engine load, increased with increasing engine speed, increased as

the spark was retarded, and showed a maximum at an equivalence ratio of 1.0,

decreasing slightly as the mixture is made leaner, and more rapidly as the

mixture is made richer.

25

The model was then used to explore the effect of engine design

variables. It was shown that changes in average temperature at cylinder

exit significantly affected the percent reacted (higher cylinder temper-

atures giving higher percent reacted). Thus increases in engine com-

pression ratio would be expected to decrease percent reacted signifi-

cantly. Increases in port wall temperature (for example through the use

of port liners) showed substantial increases in percent reacted. Increas-

ing the cross-sectional area of the port increased percent reacted by

increasing port residence time and decreasing the impact of heat

transfer. An analysis of the degree of oxidation occurring in a pipe

downstream of the port exit showed that the reaction rate steadily

decreased with increasing distance from the valve; by port exit about

two thirds of the total percent reacted had occurred. These studies

illustrated the utility of this model for improving exhaust port design,

and identified the key design variables. The percent hydrocarbon which

reacts in the port, under typical engine operating conditions, is

sufficiently large for this to be a fruitful development area.

To provide confirmation of these predicted levels of hydrocarbon

reaction in the port, and the trends with major engine operating vari-

ables, and also provide data which defines cylinder exit hydrocarbon

emission levels, an additional experimental activity was initiated. A CFR

engine set-up was developed where the exhaust gas flow could be cooled

by dilution with a cold gas stream, at a selected plane within the exhaust

port to quench the exhaust hydrocarbon oxidation reactions. By differ-

encing exhaust concentrations with and without quenching, the reduction

26

in hydrocarbons downstream of the quenching plane could be determined.

In the test set-up, the quench gas, carbon dioxide, could be injected

via a hollow-stemmed exhaust-valve through thirty-two holes drilled in

the back of the valve head to provide uniform quenching just downstream of

the valve seat, or through an injector probe which could be traversed

axially into the exhaust port along the port axis from the port exit

plane. It was shown that a quench-gas flow equal to the exhaust flow was

sufficient to quench the hydrocarbon oxidation reactions in the port.

This quench flow had no effect on indicated mean effective pressure at

typical engine operating conditions, and reduced NOx emissions by less

than ten percent.

The experimental set-up was designed, fabricated and tested within

the grant period. Since that time, a program of experiments to measure

the percent hydrocarbon reacted in the port has been completed, and the

results are described in Volume XI.

Experiments were carried out over a matrix of engine operating condi-

tions comparable to those studied by Caton (Volume IX) in the exhaust port

modelling program. It was shown that the reaction rate decreased rapidly

as the quenching plane was moved downstream of the cylinder exit, thus

confirming that the upstream portion of the exhaust port is the critical

design area. Trends and magnitudes in percent hydrocarbon reacted in the

port, with indicated mean effective pressure, speed, equivalence ratio,

and spark retard were comparable with model predictions, and values up to

40 percent were measured. That the maximum percent reacted occurs at

stoichiometric operating conditions, as predicted by the model, was con-

27

firmed. These experiments also defined the variation of cylinder exit

hydrocarbon level with indicated mean effective pressure, engine speed,

equivalence ratio, spark retard and compression ratio. The trends with

indicated mean effective pressure, speed, equivalence ratio and compression

ratio, at cylinder exit compared to exhaust system exit, are sufficiently

different for this data to be extremely valuable in future studies of

in-cylinder hydrocarbon formation mechanisms.

28

6. SUMMARY

A concise summary of the major achievements of the program is pro-

vided below.

a) Laminary Quenching Studies

1) Correlations for the flame stand-off distance for a laminarflat-flame burner, and for critical tube diameter for ex-tinguishment of a laminar flame in a tube have been developedand verified with laminar flame-quenching data. It was shownthat these correlations are relevant to flame quenching in theSI engine context.

2) A combustion bomb and method of analysis for measurement oflaminar flame speeds of fuel-air mixtures under conditionsrelevant to engines, have been developed. This facility hasbeen used to measure laminar flame speeds for propane,methanol, isooctane and indolene fuel-air mixtures underconditions appropriate to spark-ignition engine operation.The laminar flame speed characterizes the combustion chem-istry in the correlations developed for flame quenching aswell as in other engine combustion problems.

b) Quench Layer Studies in an Engine

1) An experimental technique based on the GM fast-acting valvefor efficient sampling of gases from the engine cylinder atany point in the entire engine cycle, with an acceptably lowleak rate for quench layer studies, has been developed andtested successfully.

2) A model for flow into the sampling valve, which permits anestimate of the mass of hydrocarbons per unit area of com-bustion chamber surface in the quench layer and the thicknessof the quench layer to be estimated from sampling valve data,has been developed. The model allows for viscous boundarylayer effects.

3) With these experimental and analytical techniques, estimatesof quench layer composition and thickness in an operating CFRengine have been made. At typical engine operating conditionsthe mass of hydrocarbons per unit area is about 1 g HC/cm2.The effects of variations in equivalence ratio, inlet pressure,spark advance, exhaust gas recycle and coolant temperature onquench layer properties have been quantified.

29

4) The laminar quenching theory has been applied to the enginecontext, over the above range of engine variables, and hasbeen shown to be in reasonable agreement with the quenchlayer mass of hydrocarbons per unit area determined experi-mentally.

c) Flow Visualization Studies in the Transparent Engine

1) A square cross-section transparent engine for Schlieren photo-graphy flow visualization has been designed, constructed andoperated successfully to produce spark-photographs and high-speed movies which illustrate flow features in the enginecylinder through the complete engine operating cycle.

2) Flow of denser gas out of the spark plug crevice duringexpansion, and flow of this gas out of the cylinder during theexhaust blowdown process has been observed.

3) Measurements were made from these movies of thermal boundarylayer thickness on the cylinder head and upper cylinder wall.The thickness varied over the operating cycle between 0.2 and2 mm. A substantial part of the boundary layer in theseregions was entrained by the exhaust blowdown flow.

4) A jet-type flow out of the crevice volumes between the piston,piston rings and cylinder wall late in the expansion processhas been defined. This jet penetrates to the exhaust valve(and maybe beyond) during blowdown, and is pushed out of the

cylinder during the exhaust stroke. It represents a possiblealternative mechanism to the previously proposed for crevice

hydrocarbons leaving the cylinder: the vortex formed by thepiston motion during the exhaust stroke.

5) The vortex generated in the piston-crown cylinder-wall corner

by the piston motion during the exhaust stroke was observedin the Schlieren movies, and its dimensions were shown tocorrelate with previously developed scaling laws. This vortexlifts off the piston at the end of the exhaust stroke andmoves towards the exhaust valve.

d) Studies of Heat Transfer, Mixing and HC Oxidation in the Exhaust

Port

1) Experiments were carried out in a CFR engine to measure thegas temperature at exhaust port exit, to provide data fordeveloping an exhaust-port heat-transfer model. A fine-wireresistance-thermometer was developed for this purpose.

30

2) A model for the flow in the exhaust port, which includes newcorrelations for the exhaust-port heat-transfer process,includes gas mixing due to the jet-type flow through thevalve seat at the start of the exhaust blowdown, and incor-porates a kinetic model for hydrocarbon oxidation has beendeveloped to match this gas temperature data.

3) The model has been used to define the degree of HC reactionin the exhaust port as a function of engine load, speed,spark advance and equivalence ratio. Percent HC reacted inthe exhaust port varied from 10 to 40 percent.

4) The model was used to predict the effect of changes in keydesign variables--port wall-temperature, port area, portlength--on percent HC reacted.

5) An experiment to quench the exhaust flow in a CFR enginewithin the exhaust port, to determine directly the percentHC reacted in the port, was planned. The measured percentHC reacted in the port confirmed the magnitude and trendspredicted by the model. The data also defined several enginevariables for which cylinder exit HC and exhaust exit HCshowed different trends.

31

ACKNOWLEDGMENT

The work summarized in this Volume, and described in more detail

in Volumes II to XI, has been supported by a grant from the Engine Research

Department, General Motors Research Laboratories. The participants in this

program are grateful for the valuable assistance provided by members of

General Motors research staff throughout the course of this grant.

32

APPENDIX A

HYDROCARBON FORMATION AND OXIDATION IN SPARK

IGNITION ENGINES

A research program in the Sloan Automotive Laboratory at MIT funded

by the Engine Research Department, General Motors Laboratories.

The following theses and papers have been published or are in progress:

1. Theses Completed:

B.S. Degree

John Harrington, "Visualization Studies of Flow through anIC Engine Exhaust Valve" May 1976

Masters Degree

Stephen P. Hansen, "The Design and Testing of a Flow VisualizationEngine" January 1978

Mohamad Metghalchi, "Laminar Burning Velocity of Isooctane-Air,Methane-Air, and Methanol-Air Mixtures at High Temperatureand Pressure" October 1976

Andrej K. Wroebel, "Experimental Study of Quench Layer HydrocarbonsDuring the Expansion Stroke of a Spark-Ignition Engine,Using a Fast Sampling Valve" February 1978

Joachin Sanchez-Barsse, "Schlieren Flow Visualization ofCombustion and Quench Zone Motion in an InternalCombustion Engine" March 1979

Paul Weiss, "A Parametric Study of Quench Layer Properties byRapid Sampling inc a Spark-Ignition Engine" January 1980

John Mendillo, "Quenching Studies to Determine HydrocarbonOxidation in the Exhaust Port of a Spark-Ignition Engine"

February 1980

Ph.D. Degree

Colin Ferguson, "Standoff Distances on a Flat Flame Burner"February 1977

33

Jerry Caton, "Heat Transfer, Mixing and Hydrocarbon Oxidationin an Engine Exhaust Port" September 1979

Mohamad Metghalchi, "Laminar Burning Velocity of Mixtures ofAir with Indolene, Isooctane, Methanol and Propane"

October 1979

2. Theses in Progress:

Masters Degree

Edward Lyford-Pike, "Boundary-Layer Studies in the MIT SquareCross-Section Flow Visualization Engine" 1980

Ph.D. Degree

Mehdi Namazian, "Quench Gas Motion Within the Cylinder of aSpark-Ignition Engine" 1980

3. Publications to Date

Colin R. Ferguson and James C. Keck, "Hot-wire Pyrometry,"J. Appl. Phys. 49(6), June 1978, 3031-2.

Colin R. Ferguson and James C. Keck, "Stand-Off Distances on aFlat Flame Burner," Combustion and Flame 34, (1979), 85-98.

P. Weiss, A. K. Wroebel and J.C. Keck, "A Parametric Study of QuenchLayer Hydrocarbons Using a Fast Sampling Valve," Paper No.CSS/CI 79-16 presented at Spring Meeting of Central StatesSection, The Combustion Institute, April 9-10, 1979 inColumbus, Indiana.

M. Metghalchi and J.C. Keck, "Laminar Burning Velocity of Propane-

Air Mixtures at High Temperature and Pressure," Paper No.CSS/CI-79-04 presented at Spring Meeting of Central StatesColumbus, Indiana. Submitted to Combustion and Flame.

J.A. Caton and J.B. Heywood, "Models for Heat Transfer, Mixing

and Hydrocarbon Oxidation in an Engine Exhaust Port,"SAE Paper 800290 presented at SAE Automotive Congress,February/March 1980.

M. Namazian, S. Hansen, E. Lyford-Pike, J. Sanchez-Barsse, J.B.

Heywood and J.M. Rife, "Schlieren Visualization of the Flowand Density Fields in the Cylinder of a Spark-Ignition Enginethrough the Complete Operating Cycle", SAE Paper 800044presented at SAE Automotive Congress, February/March 1980.

4. Publications in Preparation

34

J.C. Caton and J.B. Heywood, "An Experimental and Analytical Studyof Heat Transfer in an Engine Exhaust Port," to be submittedto International Journal of Heat and Mass Transfer.

J. Mendillo and J.B. Heywood, "Quenching Studies of HydrocarbonOxidation in the Exhaust Port of an SI Engine" to bepresented at SAE meeting, 1980.

35